Insights from a journey of building high-tech companies, bridging research and business, and scaling AI globally.
AI startups don’t succeed because of brilliant models alone—they succeed when technology becomes scalable, trusted, and valuable for enterprise customers. Over the last 25 years, the dynamics of startup creation and acceleration have shifted, highlighting the need for operational strength, customer focus, and stronger bridges between research and entrepreneurship.
Today we’re surfing with Andrea Ridi, high-tech entrepreneur and professor at Hult International Business School in Boston, to explore the lessons learned, the gaps that remain, and the opportunities shaping the future of AI innovation.
Before diving into this journey, make sure to be the first to access exclusive content.
Your journey as an entrepreneur of innovation began over 25 years ago, when you scaled your first start-up from Italy to a global stage, drawing the attention of a major multinational. Is this same “boldness” still what drives a start-up today?
When I think back to my first international startup, what strikes me most is the jump we made from a mathematical theorem to a real product. We had developed a new machine learning approach that could create intelligible models, and from that we built an engine.
At first, we licensed the engine to a small international company—this helped us understand how the technology worked in a real market. But the truly bold step was building a complete product on top of it and deciding to launch from Boston. The beginning was challenging, but over time, we gained recognition, even being named among the ten most disruptive startups in the world. We even caught the attention of MIT and large corporations.
Working with big corporations was the hardest part. I quickly realized that it’s not technology that matters the most to them—it’s the ability to deliver value. And to deliver value, you need scalable operations, budgets that can support growth, and a language that corporate stakeholders can understand. Technology is just an enabler; without execution and market coverage, it doesn’t translate into impact. That was the biggest lesson of my career.
Boldness remains the spark. But boldness is not only about taking risks with new technology—it is about building the operational muscle to deliver value to big customers. That is what truly makes a startup global.
Which are the major transformations you’ve seen over these years in the dynamics of start-ups origination and acceleration?
In the last few years, I’ve worked closely with many multinationals, always trying to understand the barriers between them and startups. The gaps are huge: corporates are risk-averse, they struggle with well defined budgets, where there isn’t room for innovative products, and they often speak a language that startups are not understanding.
On the other hand, startups usually cannot deliver value at large scale, making them seem incompatible with large customers. For me, this is still the main difference between the U.S. ecosystem and many others: in the U.S., startups are pushed to build operations that can really scale, turning technology into value. When I speak with international founders today, I often see this missing piece.
What changes would you like to see in how universities, research institutions, and startups collaborate to accelerate AI innovation? Do you see any differences between the US and European ecosystems?
What I would like to see is a much stronger bridge between universities, research institutions, and startups, especially in Europe. Right now, Europe produces outstanding technology, but too often it gets trapped inside layers of procedures and norms that slow down the process of turning it into impact.
The biggest difference I see with the U.S. is the attitude toward risk. In the U.S., the system is designed to embraceexploit risk: failure is accepted as part of the process, and the possibility of outsized returns keeps founders, investors, and even universities moving forward aggressively. In Europe, the instinct is to mitigate risk at every step. This makes the system safer, but it also flattens the risk–return equation—there’s no way to achieve big outcomes if you are always focused on limiting exposure.
This is not about weak research. Europe has brilliant labs, great scientists, and strong ideas. The problem is how to move faster from research to market, how to create the conditions where startups can take the leap, scale, and compete globally.
To accelerate AI innovation, we need more programs and policies that encourage bold experiments, reduce barriers to spin-offs, and connect researchers directly with entrepreneurial teams that are ready to scale internationally.
As both a scientist and entrepreneur, how do you balance the depth of rigorous research with the speed and pragmatism required in the startup world?
I was trained as a physics researcher, and the most important lesson from that training was how to deal with the unknown—especially the “unknown unknowns.” In science, you never have all the answers. You build hypotheses, design experiments to test them, analyze the results, and then create a new set of hypotheses. That cycle never ends.
For me, a startup works in exactly the same way. When you begin, you don’t know the market, you don’t know how customers will react, and there are many things you don’t even know that you don’t know. So I approach it with the same mindset: treat the startup as a living experiment. The difference is that, in startups, the experiments include not only the product but also the market, the finance, the people, and the operations.
Balancing research and entrepreneurship, then, is not about choosing between depth and speed. It’s about applying the same rigorous method to a faster, more complex environment. In research, you test theories; in startups, you test value propositions. In both cases, you move step by step through uncertainty, learning quickly, and adapting based on evidence.
Having co-founded more than 10 high-tech startups, what do you see as the most critical factors that determine whether an AI startup succeeds or fails?
From my experience, the most critical factor that decides whether an AI startup succeeds or fails is its ability to deliver value at scale. Technology alone, no matter how brilliant, is not enough. What matters is whether that technology can be transformed into real solutions that customers adopt, pay for, and can rely on globally.
For AI startups in particular, I see three layers that make the difference.
The first is problem clarity: are you solving a real pain point, not just showing off a clever model? Too many teams fall in love with the technology instead of the customer problem.
The second is operational scalability: once you have a product, can you actually deliver it at the level of reliability, integration, and support that enterprises and global customers expect? This is where many AI startups fail—they underestimate the operational muscle required.
The third is trust and risk management: AI is powerful but also sensitive. Customers need to trust how data is handled, how models behave, and how the company will respond when things go wrong. Without that trust, adoption stalls.
So, success is not defined by the model accuracy or the novelty of the algorithm. Success comes when a startup can consistently convert technology into measurable value, scale that value through solid operations, and build enough trust that customers are willing to bet on them.
That’s what separates the winners from the rest.
Looking at current AI research, which areas do you think are most underexplored but hold high potential for future breakthroughs and startup opportunities?
I believe one of the most underexplored yet high-potential areas in AI is the development of small, specialized language models—models that can capture tribal knowledge, the deep expertise inside a community, a company, or even an individual. The future, in my view, is not only about massive foundation models but about a decentralized, agentic vision of AI.
In this vision, every company and every individual will be able to create their own AI, trained to understand and express their unique value. Instead of one-size-fits-all systems, we will see a fabric of specialized agents that collaborate, exchange knowledge, and provide highly contextualized support.
This shift is powerful because it moves AI closer to the user, closer to the domain, and ultimately closer to value creation. It also aligns with the need for privacy, sovereignty, and control—organizations want to own their intelligence, not just rent it from big platforms. For startups, the opportunity is huge: building tools and infrastructures that make it simple for anyone to capture their knowledge and turn it into an intelligent, adaptive agent.
Andrea Ridi has spent over 25 years as a scientist, entrepreneur, and technology leader, deeply engaged in the international startup ecosystem. His career has been defined by a passion for bridging groundbreaking research with real-world applications, nurturing startups, and connecting diverse ecosystems.
As the founder of more than 10 high-tech startups across fields such as AI, IoT, advanced materials and software, including 4 successful exits, he recently co-founded ScaleUp Labs, a Boston-based venture studio focused on building market-first AI startups. Alongside his ventures, he has mentored over 40 international startups spanning Europe and South America, driven by a strong belief in the philosophy of giving back.
Andrea has also collaborated with universities, research centers, and industry partners worldwide. As a professor at Hult International Business School, he is committed to fostering innovation and preparing the next generation of entrepreneurs to create practical, scalable answers to today’s global challenges.
His current focus lies at the intersection of AI and entrepreneurship: the future of AI, new business models enabled by generative technologies, and the decentralization of AI ecosystems—with the broader goal of enabling transparent, resilient, and human-aligned innovation.